The data shows a disconnect. Over the past twelve months, Cadence Design Systems—a company that designs the software used to design every AI chip—has seen its stock trade at a forward P/E of roughly 35x. Meanwhile, Nvidia trades at 45x, and the broader semiconductor ETF is at 25x. The gap is narrow, but the narrative is wrong. Markets are treating Cadence as a cyclical software vendor, not what it really is: the toll collector on every AI chip that gets taped out.
I’ve been watching this for months, parsing the order flow around earnings calls. The algo traders are pricing in a slowdown in EDA spend because they think chip design cycles lag. They’re missing the structural shift. In 2022, after the Terra collapse, I coded a script to track on-chain inflows into exchanges. I learned that when the market panics, the smart money is looking at where the leverage is hiding. Today, the leverage in AI infrastructure is hiding in the assumption that design tools are a commodity. They’re not.
Context: The EDA Industry and Its Invisible Leverage Cadence and Synopsys control roughly 60% of the EDA market—the software that allows anyone to design a chip from concept to tapeout. The global EDA market is about $16-18 billion in 2024, tiny compared to the $600 billion semiconductor market. But here’s the kicker: every dollar of EDA revenue supports roughly $200-300 of semiconductor output and $5,000-10,000 of end-user tech value. This is the “superlinear leverage” that the market consistently fails to price.
The CEO of Cadence recently argued the company is undervalued amid the AI boom. Typical CEO talk, you’d think. But the timing is interesting. The interview was published on Crypto Briefing—a site that usually covers blockchain, not semiconductors. That’s a signal. The CEO is trying to reach a broader tech investor base, beyond the usual semiconductor analysts. He’s saying: “You’re measuring us with the wrong ruler.”
From my perspective as a quant trader, this is a classic mispricing of the “tool tax.” In crypto, we saw the same thing with NFT marketplaces—OpenSea’s fees were treated as a percentage of volume, but the real value was in the network effects of the platform. Cadence has a similar dynamic. Every AI chip—whether it’s Nvidia’s H100, AMD’s MI300, or a custom ASIC from Google or Amazon—requires EDA tools. The design cost for a 2nm chip is now $5-7 billion, and EDA/IP accounts for 25-30% of that. As nodes shrink, that percentage rises.
Core: The Order Flow Analysis of EDA Spending I ran a rough order-flow model based on publicly available data from Cadence’s 10-K and the earnings calls of its top customers. The top five chip design companies—Nvidia, Apple, AMD, Qualcomm, Broadcom—all have multi-year contracts with Cadence. The contract value is typically tied to the number of design seats and the complexity of the nodes used.
Here’s the key insight: the revenue per design start is increasing faster than the number of design starts. In 2020, a 7nm design cost about $500 million in EDA/IP. By 2025, a 2nm design will cost over $1.5 billion in EDA/IP. That’s a 3x increase in 5 years, driven by the complexity of AI chips with high-bandwidth memory, chiplets, and advanced packaging.
Yet, the market values Cadence at a 35x P/E, while Nvidia is at 45x. The implied growth rate for Cadence is about 12-15% CAGR. But the actual growth in design complexity is closer to 20% per node transition. The disconnect is a mispricing of the “tool tax” elasticity.
In my own trading, I’ve used this type of analysis to short volatility on EDA stocks before earnings. The pattern is consistent: the market underestimates the recurring revenue component. In 2024, when the ETH ETF was approved, I noticed that institutional desks were mispricing short-term volatility because they were using rigid risk models that ignored crypto-native signals. The same thing is happening here. The market is using a software valuation framework for a company that is becoming a platform operating system for chip design.
Contrarian: The Retail vs. Smart Money Divide The contrarian view is that Cadence is overvalued because AI chip demand will eventually slow, and EDA spending will follow. But that’s a linear extrapolation from previous cycles. The data shows that design starts are counter-cyclical. During the 2022-2023 semiconductor downturn, Cadence’s revenue grew 14% while the broader chip industry contracted. Why? Because when chip companies are in a downturn, they double down on designing new products to be ready for the next upcycle. EDA is the last thing they cut.
Retail investors often look at the headline P/E and compare it to Nvidia, missing the fact that Cadence has a higher gross margin (88% vs. Nvidia’s 70%) and a more predictable revenue stream. The smart money is slowly rotating into EDA as a defensive play on AI, but the volume isn’t there yet. I can see it in the options flow—call buying on Cadence has been tepid compared to Synopsys, which recently acquired Ansys for $350 billion in a deal that closed in 2024. The market is pricing Synopsys for a premium, but Cadence is the better value because it has a higher operating margin and a more efficient business model.
Takeaway: The Gap Between Expectation and Execution The tape will tell you that Cadence is “just” an EDA company. But the code tells a different story. Every AI chip that gets taped out leaves a trail in the design logs—a trail that leads back to Cadence’s tools. The market is pricing the expectation of a slowdown, but the execution is accelerating. If the AI infrastructure capex continues to grow at 20%+ CAGR, the “tool tax” will compound.
I’m not saying buy Cadence. I’m saying the data is clear: the valuation gap between EDA and the rest of the AI stack is a structural inefficiency that will eventually close. The ledger remembers what the code tries to hide.
Signatures: - “The ledger remembers what the code tries to hide.” - “Uptime is a promise; downtime is the truth.” - “I trade the gap between expectation and execution.”